JP2016536652A - Real-time speech evaluation system and method for mobile devices - Google Patents

Abstract

The present invention discloses a real-time speech evaluation system and method in a mobile device. The system includes a collection module (110) used for collecting voice data of voices waiting for evaluation, and an identification module (130) used for recognizing voice data collected by the collection module (110) as text data. A matching module (150) used for matching the text data obtained by recognition by the identification module (130) with the text data of the voice sample in the voice sample repository, and obtaining a matching result, a predefined evaluation policy, and Based on the matching result obtained by matching by the matching module (150), the pronunciation score of at least one character or character string in the evaluation waiting speech and / or the pronunciation score of the evaluation waiting speech is obtained and output. And an evaluation module (170) used for the above. By completing the voice evaluation system in the mobile device, not only can the dependence of the mobile device on the network be reduced, but also the user can be fed back with real-time voice evaluation, and the user experience effect can be enhanced.

Description

  The present invention relates to the field of computer technology, and more particularly, to a real-time voice evaluation system and method in a mobile device.

  Many of the conventional voice evaluation systems use a computer as a client, a user records with a microphone connected to the computer, uploads voice data to a server via a network, and evaluates by an algorithm executed on the server. The evaluation algorithm is operated on a computer of a server having a relatively large computing resource (CPU resource, memory resource, storage resource).

  With the proliferation of mobile devices, users have begun to change from computer clients to mobile device clients. When migrating evaluation system clients to mobile devices, the following solutions are often used. In other words, voice data is collected from the mobile device client, the voice data is sent to the server via the network, evaluated by the voice evaluation algorithm executed on the server, and the evaluation result is returned to the mobile device client via the network. To do.

  The conventional technology relies on the network being connected. In some aspects, sending voice data over a network is expensive, and in some aspects, mobile devices are not connected to a reliable network at any time. Both of the above two points easily give a negative user experience to the voice evaluation system, and increase the extra cost for the construction and maintenance of the voice evaluation system server.

  The present invention has been made in view of the above problems, and provides a real-time speech evaluation system and method in a mobile device in order to solve the above-mentioned problems or some problems.

  By completing the voice evaluation system on the mobile device, the voice evaluation system can not only reduce the dependency on the network, reduce the flow loss of mobile device and server message transmission, but also the user real-time voice evaluation You can also give feedback. Therefore, voice practice can be performed anytime and anywhere using the voice evaluation system, and the user experience effect can be enhanced.

  According to one embodiment of the present invention, a real-time speech evaluation system in a mobile device is provided. The system is in a collection module used to collect voice data of evaluation-waiting voice, an identification module that recognizes voice data collected by the collection module as text data, and text data and voice sample storage recognized by the identification module. Matching with text data of voice samples and using a matching module for obtaining matching results, a predefined evaluation policy and a matching result obtained by the matching module, and using at least one character or character string of voice to be evaluated And an evaluation module for obtaining and outputting a pronunciation score and / or a pronunciation score of a voice awaiting evaluation. Here, the evaluation-waiting voice includes at least one letter voice or a character string voice.

  Preferably, the system further comprises a display module used to display text data of the audio samples in the audio sample repository.

  The collection module is used to further collect voice data as an evaluation waiting voice input based on the text data of the voice sample in the voice sample store displayed by the display module.

  Preferably, the system uses the pronunciation score of the speech waiting for evaluation output from the evaluation module and / or the pronunciation score of at least one character or character string in the speech awaiting evaluation as a predefined pronunciation score threshold. If the pronunciation comparison score of the score comparison module for comparison and the speech waiting for evaluation is lower than the predefined pronunciation score threshold, the pronunciation score is determined from the predefined pronunciation score threshold among the text data displayed by the display module. A mark is applied to low text data, and / or if the pronunciation score of a character or character string is lower than a predefined pronunciation score threshold value in the voice to be evaluated, a pronunciation score is selected from the text data displayed by the display module. A marker used to mark characters or strings that are below the predefined pronunciation score threshold. -A king module.

  Preferably, the matching module further performs a matching operation on the text data recognized by the identification module with the text data of the voice sample in the voice sample repository based on the Levenshtein Distance editing distance algorithm to obtain a matching result.

  Preferably, when the predefined evaluation policy matches the text data obtained by recognition with the text data of the voice sample in the voice sample store, the character in the text data obtained by recognition based on the voice data or The a posteriori probability of the character string is defined as the pronunciation score of the character or character string in the evaluation waiting speech, and the average score of the pronunciation scores of all the characters or character strings in the evaluation waiting speech is defined as the pronunciation score of the evaluation waiting speech.

  Preferably, the system further comprises a storage module used to store the audio sample repository. However, at least one audio sample is included in the audio sample store.

  According to another embodiment of the present invention, a real-time speech evaluation method in a terminal device is further provided. It includes the steps of collecting voice data of the awaiting evaluation voice, recognizing the collected voice data as text data, matching the text data obtained by recognition with the text data of the voice sample in the voice sample warehouse, Obtaining a matching result; according to a predefined evaluation policy and the matching result, obtaining a pronunciation score of at least one character or character string in the evaluation waiting speech and / or a pronunciation score of the evaluation waiting speech; And outputting. However, the evaluation waiting voice includes at least one letter voice or a string voice.

  Preferably, prior to the step of collecting voice data of the voice waiting for evaluation, the method further comprises displaying text data of the voice sample in the voice sample.

  In response, the step of collecting the voice data of the evaluation waiting voice collects the voice data as the evaluation waiting voice input based on the text data of the voice sample in the voice sample storage displayed by the user. It is.

  Preferably, the method compares the output pronunciation score of the speech to be evaluated and / or the pronunciation score of at least one character or character string in the speech to be evaluated with a predefined pronunciation score threshold. And, if the pronunciation score of the speech awaiting evaluation is lower than a predefined pronunciation score threshold, mark the text data whose pronunciation score is lower than the predefined pronunciation score threshold from the displayed text data, and / or Alternatively, when the pronunciation score of at least one character or character string in the speech waiting for evaluation is lower than a predefined pronunciation score threshold, the character whose pronunciation score is lower than the predefined pronunciation score threshold from the displayed text data or The method further includes marking the character string.

  Preferably, the step of matching the text data obtained by the recognition with the text data of the voice sample in the voice sample store and obtaining the matching result is based on the Levenshtein Distance edit distance algorithm. Is matched with the text data of the voice sample in the voice sample store to obtain a matching result.

  In an embodiment of the present invention, voice data of an evaluation waiting voice is collected through a real-time voice evaluation system in a mobile device, the collected voice data is recognized as text data, and the text data obtained by the recognition is stored in a voice sample store. Matching with text data of a certain audio sample, a matching result is obtained. Further, based on the predefined evaluation policy and matching result, the pronunciation score of the speech awaiting evaluation and / or the pronunciation score of at least one character or character string in the speech awaiting evaluation is obtained and output. Completing the voice evaluation system at the client terminal of the mobile device not only reduces the dependence of the mobile device on the network, reduces the flow loss of message transmission of the mobile device and server, but also provides real-time voice evaluation to the user You can give feedback. Therefore, there is an effect that voice practice can be performed anytime and anywhere using the voice evaluation system.

  The above description is an outline of the present invention, and in order to clearly understand the technical means of the present invention, it may be carried out according to the contents described in the specification. Furthermore, in order to make the aforementioned and other objects, features, and advantages of the present invention easier to understand, embodiments of the present invention will be described below.

Various other advantages will become apparent to those skilled in the art upon reading the following detailed description of the preferred embodiment. The drawings are only used to illustrate preferred embodiments and are not intended to limit the invention. In each drawing, the same reference numerals denote the same parts.
It is a block diagram which shows typically the real-time audio | voice evaluation system 100 in the mobile device based on the Example of this invention. 6 is a flowchart schematically showing a real-time speech evaluation method 200 in a mobile device according to an embodiment of the present invention.

  Next, exemplary embodiments of the present disclosure will be described in more detail with reference to the drawings. While illustrative examples of the present disclosure are shown with figures, it should be understood that the present invention may be implemented in various ways, regardless of the scheme described in the present disclosure. Conversely, these examples are provided in order to provide a better understanding of the present disclosure and to fully convey the technical scope disclosed herein to those skilled in the art.

  It should be understood that a person skilled in the art can find the realization of each structure within the spirit, principle, and scope of the present invention included in the present invention without any clear explanation or description in the present specification.

  All examples and conditional languages described herein are for illustrative and teaching purposes only and are intended to give the reader a deep understanding of the principles and concepts that contribute to the prior art of the inventor. Thus, the present invention is not limited to these specific examples and conditions.

  Any interpretation or description of the principles, aspects, and examples or specific examples of the invention described herein is meant to include all equivalents or equivalents in structure and function thereof. Also, such equivalents or equivalents should include known and future developed equivalents or equivalents, i.e., any structure can carry out the development of the same function. It is.

  Those skilled in the art should understand that the block diagram displayed in the accompanying drawings is a schematic diagram showing a structure or a circuit for realizing the present invention. Similarly, any flowcharts and the like displayed in the drawings attached to the specification show the processes actually executed by various computers or processors, and whether or not these computers or processors are clearly shown in the drawings is related. Should not be understood.

  Within the scope of the claims, use to perform a module with limited function includes any embodiment for performing the function, for example (a) a combination of circuit capacitors of the function, or ( b) Including execution of any form of software, including firmware and microcode, etc., used in combination with appropriate circuitry to execute the software that implements the function. It should be understood that any module / part or capacitor capable of providing these functions is equivalent to a module defined in the claims by combining the functions provided in each module with the claimed embodiments of the claims. is there.

  The term “example” in the specification means that the features, structures, and the like described in accordance with the example are included in at least one embodiment of the present invention, and therefore appear in various places in the specification. The coming term “in the embodiment” does not necessarily refer to the same embodiment.

  As shown in FIG. 1, based on the real-time speech evaluation system 100 in the mobile device according to the embodiment of the present invention, a collection module 110, an identification module 130, a matching module 150, and an evaluation module 170 can be mainly provided. It should be understood that the connection relationships of the modules shown in FIG. 1 are only examples, and engineers in this field can adopt other connection relationships. It is only necessary to realize the function.

  In this specification, the function of each module can be realized by using dedicated hardware or using hardware that performs processing in combination with appropriate software. Such hardware or dedicated hardware can include a dedicated integrated circuit (ASIC), various other circuits, various processors, and the like. When implemented by a processor, the functions can be provided by dedicated processors, individually shared processors, or multiple independent processors, some of which may be shared. Also, it should not be understood that the processor is just hardware capable of executing software, and is a read-only storage ROM (ROM) / random access memory (RAM) for storing hardware and software that is not limited to a digital signal processor (DSP). ) And non-volatile storage devices are implicitly provided.

  According to an embodiment of the present invention, the collection module 110 is used to collect voice data of voices waiting for evaluation, and includes voices of at least one character or character strings in the voices waiting for evaluation. Preferably, the evaluation-waiting speech includes one or more combinations of Chinese words, English words, and Arabic numerals, and it should be understood that the language type of the evaluation-waiting speech is an embodiment of the present invention. It is not limited to.

  In the embodiment of the present invention, the collection module 110 plays a role of registering the evaluation waiting voice and storing the voice data of the evaluation waiting voice. Preferably, the collection module 110 may be a conventional microphone, and the user inputs an awaiting evaluation voice to the system 100 through the microphone. For example, the content of the awaiting evaluation voice may be the following English “Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo.”. Preferably, the system 100 converts and stores the voice data of the voice waiting for evaluation into a voice file in the .wav format via the collection module 110. That WAV format is an audio waveform file format. It should be understood that the embodiments of the present invention are not limited to the specific structure of the collection module 110.

  According to an embodiment of the present invention, the identification module 130 is used to recognize the voice data collected by the collection module 110 as text data.

  That is, the voice data of the evaluation waiting voice described in the above example can be recognized as the following text data WELCOME TO LIU LI SHUO! MY NAME IS PETER.

  Preferably, in the embodiment of the present invention, the speech recognition model employed by the identification module 130 is a hidden Markov model (HMM) having a mixed Gaussian distribution as an output probability distribution.

  The identification module 130 performs fixed point calculation and recognizes the voice data collected by the collection module 110 as text data. For example, the fixed point calculation is performed by the following method. Of course, it is not limited to this.

Method 1, conventional speech recognition algorithm has many floating point operations, fixed point DSP (fixed point DSP is completed by integer operation or decimal point operation, data format does not include data code, and normal fixed point DSP is 16 Can be used to perform floating point operations, and a number scaling method is used to convert a floating point number to a fixed point number. The number scaling method is to determine the position of the decimal point at a fixed point. Q display method is a regular scaling method, the display mechanism is fixed point number x and floating point number is y, and the conversion relationship between fixed point number and floating point number in Q display method is to convert floating point number y to fixed point number x then, it is x = (int) y × 2 Q.

  Method 2, (1) Define and simplify the algorithm structure. (2) Determine key variables in functions that require quantization. (3) Collect statistical information of key variables. (4) Determine the correct display of key variables. (5) Determine the fixed point format for other variables.

  Thus, in the embodiment of the present invention, a general floating point can be replaced by using a fixed point operation, and the general floating point is replaced by an integer to represent the output probability of the recognition result. Since the fixed point operation can be used in the embodiment of the present invention, and the fixed point operation does not need to define many parameters for the floating point operation, the identification module 130 has few system resources (CPU resource, memory resource, storage). The recognition process can be completed under the circumstances of occupying resources. It should be understood that the embodiment of the present invention is not limited to the specific type of recognition model that the identification module 130 employs for character recognition.

  According to the embodiment of the present invention, the matching module 150 matches the text data obtained by the recognition module 130 with the text data of the voice sample stored in the voice sample repository, and obtains a matching result.

  Preferably, in the embodiment of the present invention, the text data of the audio sample in the audio sample store may be text data previously stored in the audio sample store. For example, the following text data “WELCOME TO LIU LI SHUO! Save NAME IS PETER. I'M AN ENGLISH TEACHER AT LIU LI SHUO in the audio sample store.

  Preferably, in the embodiment of the present invention, the matching module 150 further performs a matching operation on the text data obtained by the recognition module 130 based on the Levenshtein Distance edit distance algorithm and the text data of the voice sample in the voice sample repository. To obtain matching results. Among them, the matching result is obtained by matching the text data obtained by the recognition module 130 with the text data of the voice sample in the voice sample store, and the text data obtained by the recognition module 130 by using the voice sample store. May not be matched with the text data of the audio sample in It should be understood that the present invention is not limited to the matching algorithm in which the matching module 150 is employed in the embodiment of the present invention.

  According to the embodiment of the present invention, the evaluation module 170 generates the pronunciation of at least one character or character string in the evaluation-waiting speech based on the predefined evaluation policy and the matching result obtained by the matching module 150 matching. It is used to obtain and output a score and / or a pronunciation score of a voice waiting for evaluation.

  Preferably, in the embodiment of the present invention, when the predefined evaluation policy matches the text data obtained by recognition with the text data of the voice sample in the voice sample store, The a posteriori probability of the character or character string is set as the pronunciation score of the character or character string in the evaluation waiting speech, and the average score of the pronunciation scores of all the characters or character strings in the evaluation waiting speech is set as the pronunciation score of the evaluation waiting speech.

  Preferably, in the embodiment of the present invention, the pronunciation score of the character or character string is p × based on the fact that the posterior probability of the character or character string obtained by recognition of the speech data is p (0 to 1). 100.

  Taking the English sentence listed above as an example, the overall English sentence “Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo.” The pronunciation score of each word in the above English sentence can be obtained. That is, in the embodiment of the present invention, a unigram language model composed of sentences and words may be used.

  According to an embodiment of the present invention, the real-time voice evaluation system 100 in the mobile device may further include one or a plurality of option modules, thereby realizing additional or additional functions. However, these optional modules are not necessarily indispensable for realizing the object of the present invention. According to the embodiment of the present invention, the real-time speech evaluation system 100 in a mobile device does not include these optional modules. The object of the present invention can be realized. Even if these optional modules are not displayed in FIG. 1, the connection relationship between the modules can be easily found by a person skilled in the art by the following teaching.

  Preferably, in an embodiment of the present invention, the system 100 further comprises a display module used to display the text data of the audio sample in the audio sample repository, for example, the following English sentence “Welcome to Liu Li shuo! My name” is Peter. I'm an English teacher at Liu Li shuo. "

  In response to this, the collection module 110 is further used to collect voice data as an evaluation-waiting voice input based on the text data of the voice sample in the voice sample store displayed on the display module by the user.

  That is, the collection module 110 collects voice data of the following English sentence “Welcome to Liu Li shuo! My name is Peter. I ’m an English teacher at Liu Li shuo.” Read by the user.

Preferably, in an embodiment of the present invention, the system 100 further comprises a score comparison module and a marking module, of which
The score comparison module compares the pronunciation score of the evaluation waiting speech output by the evaluation module 170 and / or the pronunciation score of at least one character or character string in the evaluation standby speech with a predefined pronunciation score threshold. In use, preferably, the predefined pronunciation score threshold can be set to 60 scores, and it should be understood that it is not limited to that particular value in the embodiments of the present invention.

  The marking module marks text data whose pronunciation score is lower than the predefined pronunciation score threshold from among the text data displayed by the display module when the pronunciation score of the speech waiting for evaluation is lower than the predefined pronunciation score threshold. And / or when the pronunciation score of at least one character or character string in the waiting voice for evaluation is lower than a predefined pronunciation score threshold, the pronunciation score having a predefined pronunciation score from the text data displayed by the display module Used to mark a character or character string below the score threshold.

  Taking the English sentence listed above as an example, if the score comparison module compares and finds that the pronunciation score of "Welcome" is lower than the predefined pronunciation score threshold, mark "Welcome" in the entire English sentence The “Welcome” color is preferably set to red.

  Preferably, in an embodiment of the present invention, the system 100 further comprises a storage module used to store the audio sample store. However, the audio sample store includes at least one audio sample, for example, the following audio sample "Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo."

  Through the embodiments of the present invention, not only reducing the dependency of the mobile device on the network and reducing the flow loss of message transmission of the mobile device and the server by completing the voice evaluation system at the client terminal of the mobile device, In addition, real-time voice evaluation feedback can be provided to the user. Therefore, there is an effect that voice practice can be performed anytime and anywhere using the voice evaluation system.

  According to another embodiment of the present invention, corresponding to the above-described real-time voice evaluation system 100 in a mobile device according to the embodiment of the present invention, the present invention further provides a real-time voice evaluation method 200 in a terminal device.

  FIG. 2 is a flowchart schematically showing a real-time speech evaluation method 200 in a mobile device according to an embodiment of the present invention. As shown in FIG. 2, the method 200 includes steps S210, S230, S250, and S270, and the method 200 starts from step S210, in which voice data of voices waiting for evaluation is collected. The voice waiting for evaluation includes voice of at least one character or character string, and preferably one or more combinations of Chinese words, English words, Arabic numerals in the voice waiting for evaluation It should be understood that the present invention is not limited to the language type of the voice waiting for evaluation in the embodiment of the present invention.

  Preferably, a user can input an evaluation waiting voice to the system 100 through a microphone. For example, the content of the awaiting evaluation voice may be the following English “Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo.” Preferably, the system 100 converts and stores the voice data of the voice waiting for evaluation into a voice file in the .wav format via the collection module 110. That WAV format is an audio waveform file format.

  Thereafter, in step S230, the collected voice data is recognized as text data. That is, through step S230, the voice data of the voice waiting for evaluation described in the above example is recognized by the following text data WELCOME TO LIU LI SHUO! MY NAME IS PETER.

  Preferably, in the embodiment of the present invention, the employed speech recognition model is a hidden Markov model (HMM) having a mixed Gaussian distribution as an output probability distribution. In other words, in the embodiment of the present invention, a general floating point is substituted by using a fixed point operation, and the general floating point is substituted by an integer number to represent the output probability of the recognition result. It should be understood that the embodiment of the present invention is not limited to the specific type of recognition model employed by character recognition.

  Thereafter, in step S250, the recognized text data is matched with the text data of the voice sample in the voice sample box, and a matching result is obtained.

  Preferably, in the embodiment of the present invention, the text data of the audio sample in the audio sample store may be text data previously stored in the audio sample store. For example, the following text data WELCOME TO LIU LI SHUO! MY NAME IS Save PETER. I'M AN ENGLISH AT LIU LI SHUO in the audio sample store.

  Preferably, in the embodiment of the present invention, in step S250, the text data obtained by recognizing based on the Levenshtein Distance editing distance algorithm is subjected to matching operation with the text data of the voice sample in the voice sample warehouse, and matching is performed. Get results. For example, the matching result includes not matching the text data obtained by recognition with the text data of the voice sample in the voice sample store. It should be understood that the embodiments of the present invention are not limited to matching algorithms.

  Thereafter, in step S270, based on a predefined evaluation policy and the matching result, obtain at least one character or string-like pronunciation score in the evaluation pending speech and / or a pronunciation score of the evaluation pending speech, and Output.

Preferably, in the embodiment of the present invention, when the predefined evaluation policy matches the text data obtained by recognition with the text data of the voice sample in the voice sample store, The a posteriori probability of the character or character string is set as the pronunciation score of the character or character string in the evaluation waiting speech, and the average score of the pronunciation scores of all the characters or character strings in the evaluation waiting speech is set as the pronunciation score of the evaluation waiting speech.
Preferably, in the embodiment of the present invention, the pronunciation score of the character or character string is p × based on the fact that the posterior probability of the character or character string obtained by recognition of the speech data is p (0 to 1). 100.

  Taking the English sentence listed above as an example, the pronunciation score of the entire English sentence “Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo.” And / or the above English sentence through step S270 The pronunciation score of each word in can be obtained. That is, in the embodiment of the present invention, a unigram language model composed of sentences and words may be used.

  According to an embodiment of the present invention, the real-time speech evaluation method 200 in the mobile device can realize additional or additional functions by further including one or more optional steps. However, these optional steps are not necessarily indispensable for realizing the object of the present invention. According to the embodiment of the present invention, the real-time speech evaluation method 200 in a mobile device does not include these optional steps. The object of the present invention can be realized. Even if these optional steps are not displayed in FIG. 2, the order of execution between the above steps can be easily found by a person skilled in the art by the following teaching. It should be pointed out that these optional steps and the order of execution of these steps can be selected according to actual needs unless otherwise specified.

  Preferably, the method 200 further includes the step of displaying text data of the text data of the voice sample in the voice sample repository, for example, the following English sentence “Welcome to Liu Li shuo! My name is Peter. teacher at Liu Li shuo. "is displayed.

  In response to this, the step (S210) of collecting the voice data of the evaluation waiting voice collects the voice data as the evaluation waiting voice inputted based on the voice sample in the displayed voice sample storage. .

  That is, voice data of the following English sentence “Welcome to Liu Li shuo! My name is Peter. I'm an English teacher at Liu Li shuo” read by the user through step S210 is collected.

  Preferably, the method 200 further includes the step of comparing the pronunciation score of the output waiting speech and / or the pronunciation score of at least one character or character string in the waiting speech to a predefined pronunciation score threshold. . Preferably, the predefined pronunciation score threshold is set to 60 scores. It should be understood that the embodiments of the present invention are not limited to specific values.

  If the pronunciation score of the speech awaiting evaluation is lower than a predefined pronunciation score threshold, mark the text data whose pronunciation score is lower than the predefined pronunciation score threshold from the displayed text data, and / or When the pronunciation score of at least one character or character string in the evaluation waiting speech is lower than a predefined pronunciation score threshold, the character or character string whose pronunciation score is lower than the predefined pronunciation score threshold among the displayed text data Mark the.

  Using the English sentences listed above as an example, if it is found that the pronunciation score of "Welcome" is lower than the predefined pronunciation score threshold, "Welcome" may be marked from the entire English sentence Preferably, the color of “Welcome” is set to red.

  Since the embodiment according to each method corresponds to the embodiment according to each device, the embodiment of each method will not be described in detail.

  In this specification, a number of specific details have been described. However, it should be understood that embodiments of the invention may be practiced without these detailed descriptions. In some embodiments, well-known methods, structures and techniques have not been shown in detail in order not to confuse the reader with an understanding of the principles herein.

  It should be understood by those skilled in the art that the modules in the devices in each embodiment can be appropriately changed and that they can be installed in one or more devices different from the embodiments. It can be done. Some modules in the embodiments can be combined into one module or nut or combination, and they can be divided into multiple sub-modules or subunits or sub-combinations. Except for the case of features and / or repulsive processes, any combination can be employed and all steps of any method disclosed herein and all modules of any apparatus can be combined. Unless otherwise stated explicitly, each feature disclosed in this document may be replaced with alternative features for similar, equivalent, and similar purposes provided.

  Embodiments in each apparatus of the present invention can be realized by hardware, can be realized by a software module executed on one or a plurality of processors, or a combination thereof. It should be understood by those skilled in the art that a part or all of functions of some or all of the modules in the embodiment of the present invention can be realized by using a microprocessor or a digital signal processor (DSP). It is. The present invention can also implement apparatus programs for executing the methods described herein (eg, computer programs and computer program products).

  It should be noted that the above examples are illustrative of the invention and are not intended to limit the invention. Those skilled in the art will be provided with various alternative embodiments, provided that they do not depart from the scope of the appended claims. In the claims, the order in which features are arranged does not mean a specific order of features. In particular, the order of the steps in a method claim does not mean that the steps should be performed in that order. Conversely, these steps may be performed in any suitable order. Similarly, the execution order of the modules in the device claim should not be limited by the order of the modules in the claim, and may be executed in any appropriate order. In the claims, you should not understand any reference in parentheses as a limitation to the claims. The term “comprising” does not exclude the presence of modules or steps that are not listed in a claim. A “1” or “one” before the term module or step does not exclude the presence of a plurality of such modules or steps. The present invention can be implemented by hardware including slightly different modules or by a suitably programmed computer or processor. Within the device claim that enumerates some modules, some of these modules can be implemented via the same hardware module. The use of the terms “first”, “second”, “third”, etc. does not indicate any order, and these terms can be interpreted as names. The terms “connection”, “coupler”, etc., as used herein, are defined as connections that can be operated in any expected form. For example, the connection is performed by a method such as software, hardware, etc., mechanically, electronically, digitally, simulationally, directly or indirectly.

Claims (10)

A real-time voice evaluation system (100) in a mobile device,
A collection module (110) used to collect voice data of voices waiting for evaluation;
An identification module (130) used for recognizing voice data collected by the collection module (110) as text data;
A matching module (150) used for matching the text data obtained by recognition by the identification module (130) with the text data of the voice sample in the voice sample repository and obtaining a matching result;
Based on a predefined evaluation policy and a matching result obtained by matching by the matching module (150), a pronunciation score of at least one character or character string in the evaluation waiting speech and / or the evaluation waiting speech An evaluation module (170) used to obtain and output a pronunciation score;
The real-time speech evaluation system in a mobile device, wherein the evaluation-waiting speech includes at least one character or speech of a character string. The system further comprises a display module used to display text data of audio samples in the audio sample repository,
The collection module (110) is further used to collect voice data as a voice to be evaluated, which is input based on a voice sample in a voice sample storage displayed on the display module by a user. The real-time audio | voice evaluation system in the mobile device of Claim 1. The pronunciation score of the evaluation pending speech output by the evaluation module (170) and / or the pronunciation score of at least one character or character string in the evaluation pending speech is used to compare with a predefined pronunciation score threshold. A score comparison module;
If the pronunciation score of the speech waiting for evaluation is lower than a predefined pronunciation score threshold, mark the text data whose pronunciation score is lower than the predefined pronunciation score threshold from the text data displayed by the display module, and / or Or
When the pronunciation score of the character or character string in the voice waiting for evaluation is lower than a predefined pronunciation score threshold, the character or character whose pronunciation score is lower than the predefined pronunciation score threshold among the text data displayed by the display module The real-time speech evaluation system for a mobile device according to claim 2, further comprising a marking module used to mark the column.   Based on the Levenshtein Distance edit distance algorithm, the matching module (150) performs a matching operation on the text data obtained by the identification module (130) and the text data of the voice sample in the voice sample store, and performs a matching result. The real-time speech evaluation system in a mobile device according to claim 1, further used to obtain   When the predefined evaluation policy matches the text data obtained by recognition with the text data of the voice sample in the voice sample store, the character or the character string in the text data obtained by recognition based on the voice data The a posteriori probability is a pronunciation score of a character or character string in the speech waiting for evaluation, and an average score of the pronunciation scores of all characters or character strings in the speech awaiting evaluation is used as the pronunciation score of the speech waiting for evaluation. The real-time audio | voice evaluation system in the mobile device as described in any one of Claims 1 thru | or 4.   5. The storage device according to claim 1, further comprising a storage module used to store the sound sample store, wherein the sound sample store includes at least one sound sample. A real-time voice evaluation system for the described mobile device. A real-time voice evaluation method (200) in a terminal device,
Collecting voice data of voice to be evaluated (S210);
Identifying the collected voice data as text data (S230);
Matching the text data obtained by the recognition with the text data of the voice sample in the voice sample warehouse and obtaining a matching result (S250);
Obtaining and outputting at least one character or string-like pronunciation score in the evaluation pending speech and / or the pronunciation score of the evaluation pending speech based on a predefined evaluation policy and the matching result ( S270),
A real-time voice evaluation method in a terminal device, comprising: a voice of at least one character or a voice of a character string in the evaluation waiting voice. Prior to the step (S210) of collecting voice data of the voice waiting for evaluation, the method further includes displaying text data of the voice sample in the voice sample;
The step (S210) of collecting voice data of the evaluation waiting voice includes
8. The real-time voice evaluation method in a terminal device according to claim 7, wherein the voice data is input as voices waiting for evaluation, which are input based on voice samples in the voice sample box displayed by the user. . Comparing the output pronunciation score of the speech to be evaluated and / or the pronunciation score of at least one character or character string in the speech to be evaluated to a predefined pronunciation score threshold;
Marking the text data whose pronunciation score is lower than the predefined pronunciation score threshold from among the displayed text data when the pronunciation score of the speech waiting for evaluation is lower than the predefined pronunciation score threshold; and / or When the pronunciation score of at least one character or character string in the waiting voice for evaluation is lower than a predefined pronunciation score threshold, the character or character whose pronunciation score is lower than the predefined pronunciation score threshold among the displayed text data 9. The method for evaluating real-time speech in a terminal device according to claim 8, further comprising the step of marking the column. The step of matching the text data obtained by the recognition with the text data of the voice sample in the voice sample warehouse and obtaining a matching result,
10. The text data obtained by recognizing based on the Levenshtein Distance editing distance algorithm is matched with the text data of the voice sample in the voice sample box to obtain a matching result. The real-time audio | voice evaluation method in the terminal device as described in any one.